Damn. Sorry about that. The argument involves integrating expressions of the form p (delta-x)/x; you wind up looking at the mathematical expectation of the logarithm of (wealth after)/(wealth before), and then exponentiating.